search

rowanz / neural-motifs

Code for Neural Motifs: Scene Graph Parsing with Global Context (CVPR 2018)
https://rowanzellers.com/neuralmotifs
MIT License
525 stars 115 forks source link

about nl_obj & nl_edge setting #91

Open fi94 opened 4 years ago

fi94 commented 4 years ago

I got error when I run './scripts/train_models_sgcls.sh 2'

the error is ''' ./scripts/train_models_sgcls.sh: line 21: 24548 Segmentation fault (core dumped) python models/train_rels.py -m sgcls -model motifnet -order left right -nl_obj 2 -nl_edge 4 -b 6 -clip 5 -p 100 -hidden_dim 512 -pooling_dim 4096 -lr 1e-3 -ngpu 1 -ckpt checkpoints/vgdet/vg-23.tar -save_dir checkpoints/motifnet2 -nepoch 50 -use_bias ''' when I set nl_obj & nl_edge = 0, the scripts is worked so, I don't know what is wrong with the setting. Can someone help me with the problem?

Iceland-Leo commented 4 years ago

I got error when I run './scripts/train_models_sgcls.sh 2'

the error is ''' ./scripts/train_models_sgcls.sh: line 21: 24548 Segmentation fault (core dumped) python models/train_rels.py -m sgcls -model motifnet -order left right -nl_obj 2 -nl_edge 4 -b 6 -clip 5 -p 100 -hidden_dim 512 -pooling_dim 4096 -lr 1e-3 -ngpu 1 -ckpt checkpoints/vgdet/vg-23.tar -save_dir checkpoints/motifnet2 -nepoch 50 -use_bias ''' when I set nl_obj & nl_edge = 0, the scripts is worked so, I don't know what is wrong with the setting. Can someone help me with the problem?

You set default values can you run on your machine? I tried, but it took up a lot of CPU memory during training (more than 64G), and the program being killed by the system. Do you have a problem?

fi94 commented 4 years ago

I got error when I run './scripts/train_models_sgcls.sh 2' the error is ''' ./scripts/train_models_sgcls.sh: line 21: 24548 Segmentation fault (core dumped) python models/train_rels.py -m sgcls -model motifnet -order left right -nl_obj 2 -nl_edge 4 -b 6 -clip 5 -p 100 -hidden_dim 512 -pooling_dim 4096 -lr 1e-3 -ngpu 1 -ckpt checkpoints/vgdet/vg-23.tar -save_dir checkpoints/motifnet2 -nepoch 50 -use_bias ''' when I set nl_obj & nl_edge = 0, the scripts is worked so, I don't know what is wrong with the setting. Can someone help me with the problem?

You set default values can you run on your machine? I tried, but it took up a lot of CPU memory during training (more than 64G), and the program being killed by the system. Do you have a problem?

I can't run on my machine when I set default( nl_obj = 1 & nl_edge = 2) values too. Does that mean I can only run this script on more than 64G CPU. And my cpu memory is 32 G. However, when I monitor the CPU memory while running the script, I didn't find that CPU usage was full.

Iceland-Leo commented 4 years ago

I got error when I run './scripts/train_models_sgcls.sh 2' the error is ''' ./scripts/train_models_sgcls.sh: line 21: 24548 Segmentation fault (core dumped) python models/train_rels.py -m sgcls -model motifnet -order left right -nl_obj 2 -nl_edge 4 -b 6 -clip 5 -p 100 -hidden_dim 512 -pooling_dim 4096 -lr 1e-3 -ngpu 1 -ckpt checkpoints/vgdet/vg-23.tar -save_dir checkpoints/motifnet2 -nepoch 50 -use_bias ''' when I set nl_obj & nl_edge = 0, the scripts is worked so, I don't know what is wrong with the setting. Can someone help me with the problem?

You set default values can you run on your machine? I tried, but it took up a lot of CPU memory during training (more than 64G), and the program being killed by the system. Do you have a problem?

I can't run on my machine when I set default( nl_obj = 1 & nl_edge = 2) values too. Does that mean I can only run this script on more than 64G CPU. And my cpu memory is 32 G. However, when I monitor the CPU memory while running the script, I didn't find that CPU usage was full.

I don't know which ones are occupying memory, maybe we can add QQ discussion (mine: 785091715).

fi94 commented 4 years ago

I got error when I run './scripts/train_models_sgcls.sh 2' the error is ''' ./scripts/train_models_sgcls.sh: line 21: 24548 Segmentation fault (core dumped) python models/train_rels.py -m sgcls -model motifnet -order left right -nl_obj 2 -nl_edge 4 -b 6 -clip 5 -p 100 -hidden_dim 512 -pooling_dim 4096 -lr 1e-3 -ngpu 1 -ckpt checkpoints/vgdet/vg-23.tar -save_dir checkpoints/motifnet2 -nepoch 50 -use_bias ''' when I set nl_obj & nl_edge = 0, the scripts is worked so, I don't know what is wrong with the setting. Can someone help me with the problem?

You set default values can you run on your machine? I tried, but it took up a lot of CPU memory during training (more than 64G), and the program being killed by the system. Do you have a problem?

I can't run on my machine when I set default( nl_obj = 1 & nl_edge = 2) values too. Does that mean I can only run this script on more than 64G CPU. And my cpu memory is 32 G. However, when I monitor the CPU memory while running the script, I didn't find that CPU usage was full.

I don't know which ones are occupying memory, maybe we can add QQ discussion (mine: 785091715).

ok I add you now

haoopan commented 4 years ago

@Iceland-Leo Have you solved this problem? Please tell me, Thx:)